GCB Bioenergy (2010), doi: 10.1111/j.1757-1707.2010.01050.

A comparison of canopy evapotranspiration for maize and two perennial grasses identied as potential bioenergy crops
G E O R G E C . H I C K M A N *, A N D Y VA N L O O C K E w , F R A N K G . D O H L E M A N * and C A R L J . B E R N A C C H I *z *Department of Plant Biology, University of Illinois, Urbana, IL, USA, wDepartment of Atmospheric Sciences, University of Illinois, Urbana, IL, USA, zGlobal Change and Photosynthesis Research Unit, Agricultural Research Service, USDA, Urbana, IL, USA

Abstract
In the Midwestern US, perennial rhizomatous grasses (PRGs) are considered one of the most promising vegetation types to be used as a cellulosic feedstock for renewable energy production. The potential widespread use of biomass crops for renewable energy production has sparked numerous environmental concerns, including the impacts of land-use change on the hydrologic cycle. We predicted that total seasonal evapotranspiration (ET) would be higher for PRGs relative to maize resulting from higher leaf area and a prolonged growing season. We further predicted that, compared with maize, higher aboveground biomass associated with PRGs would offset the higher ET and increase water-use efciency (WUE) in the context of biomass harvests for liquid biofuel production. To test these predictions, ET was estimated during the 2007 growing season for replicated plots of Miscanthus giganteus (miscanthus), Panicum virgatum (switchgrass), and Zea mays (maize) using a residual energy balance approach. The combination of a 25% higher mean latent heat ux (kET) and a longer growing season resulted in miscanthus having ca. 55% higher cumulative ET over the growing season compared with maize. Cumulative ET for switchgrass was also higher than maize despite similar seasonal-mean kET. Based on total harvested aboveground biomass, WUE was ca. 50% higher for maize relative to miscanthus; however, when WUE calculated from only maize grain biomass was compared with WUE calculated from miscanthus harvested aboveground biomass, this difference disappeared. Although WUE between maize and miscanthus differed postsenescence, there were no differences in incremental WUE throughout the growing season. Despite initial predictions, aboveground biomass for switchgrass was less than maize; thus WUE was substantially lower for switchgrass than for either maize scenario. These results indicate that changes in ET due to large-scale implementation of PRGs in the Midwestern US would likely inuence local and regional hydrologic cycles differently than traditional row crops.
Keywords: agriculture, biofuels, evapotranspiration, perennial rhizomatous grasses, water-use efciency

Received 12 March 2010 and accepted 1 May 2010

Introduction
Growing population, rising fuel demand, and the link between fossil fuel emissions and global change have spurred interest in expanding bioenergy production to reduce fossil fuel demand and mitigate climate change
Correspondence: C. J. Bernacchi, 193 ERML, 1201 W. Gregory Drive, Urbana, IL 61801, USA, tel. 1 1 217 333 8048, e-mail: bernacch@illinois.edu

(Sims et al., 2006; Tan et al., 2008). In the United States, corn ethanol has served as a rst step towards reducing dependence on fossil fuels and meeting environmental goals. However, agricultural input demands (e.g. fertilizer, pesticides), relatively low production efciency (Hamelinck et al., 2005), a perceived link between ethanol production and increased food prices (Msangi et al., 2006), and greenhouse gas emissions caused by indirect land-use change (Searchinger et al., 2008) have
1

Published 2010 This article is a US Government work and is in the public domain in the USA

2 G . C . H I C K M A N et al.
imposed limits on the benets of producing ethanol from grain. Thus, development of a meaningful modern bioenergy industry will necessitate a transition from grain fermentation and the development of advanced technology for the conversion of cellulose to fuel. Advanced cellulosic fuel production technologies are predicted to yield more fuel per unit land area with fewer inputs and fewer environmental impacts compared with grain-based ethanol production (Heaton et al., 2004; Hamelinck et al., 2005; Farrell et al., 2006; Sims et al., 2006). The perennial rhizomatous grasses (PRGs) Panicum virgatum L. (switchgrass) and Miscanthus X. giganteus (miscanthus) are two candidate feedstocks currently being developed for use in bioenergy production (Heaton et al., 2008; Sanderson & Adler, 2008). These feedstocks have emerged as promising second generation bioenergy crops because they possess many of the traits necessary for sustainable biomass production (Lewandowski et al., 2003; Heaton et al., 2004; Sanderson & Adler, 2008). Despite advantages over traditional grain-based methods of producing liquid fuels, the development of a cellulosic fuels industry would generate a large demand for biomass that would in turn result in land-use changes (McLaughlin & Walsh, 1999; Gurgel et al., 2007; De Fraiture et al., 2008; Searchinger et al., 2008). While large-scale planting of PRGs has the capacity for increased fuel production, it also suggests that there will be changes to these managed ecosystems, particularly with regard to the hydrologic cycle (Stephens et al., 2001; Hall, 2003; De Fraiture et al., 2008). The root systems of perennials penetrate deep into the soil and generally have access to deeper soil moisture than annual crops (Neukirchen et al., 1999; Stephens et al., 2001; Hall, 2003). Leaf area index (LAI) is higher for PRGs relative to corn (Dohleman & Long, 2009), providing greater area for photosynthesis and transpiration. These physiologically based differences between PRGs and Zea mays (maize) are further accentuated by a much longer growing season, differences in canopy architecture, and increased residue accumulation in the PRGs (Heaton et al., 2004; Dohleman & Long, 2009). Thus, it is critical to assess changes in evapotranspiration (ET), the combination of both ecosystem evaporation and transpiration, to assess the impacts of bioenergy crop growth on local and regional hydrology. Our primary research objective was to estimate differences in water use among maize, miscanthus, and switchgrass. We predicted that the physiological, morphological, and phenological characteristics of PRGs will result in more water used during the growing season compared with maize. However, given the anticipated higher productivity of the PRGs, we also predicted that water-use efciency (WUE) will be greater for PRGs relative to maize. These predictions were tested using micrometeorological measurements over established stands of switchgrass, miscanthus, and maize over one growing season in 2007. This experiment provides the rst ever side-by-side comparison of ET among these three species over a complete growing season.

Materials and methods

Site description
Three replicate plots each of miscanthus and P. virgatum L. cv. Cave-in-Rock (switchgrass) were established in the South Farms at the University of Illinois in 2004 and 2005 as described previously (Heaton et al., 2008; Dohleman & Long, 2009). Maize cv. 34H35 (Pioneer Hi-Bred International, Johnston, Iowa, USA) was planted at 0.76 m row spacing on May 8, 2007 and emerged on May 11, 2007. The same area had been planted with soybean in 2006 and 168 kg [N] ha1 was applied before the planting of the maize crop, following typical agronomic practices in this region of the Midwestern USA. No fertilizer was applied to either maize or switchgrass in 2006 or 2007. The plots measured ca. 30 m 60 m (0.2 ha) and consisted of mature homogenous plantings of each species planted on a FlanaganDrummer silty clay loam (ne-silty, mixed, mesic Typic Endoaquolls). The plots were planted in a randomized complete block design (n 53) and were surrounded on all sides by production-scale plantings of maize.

Micrometeorology
Given the relatively small plot sizes, ET was best estimated as the residual in the energy balance equation (Huband & Monteith, 1986; Jackson et al., 1987; Triggs et al., 2004): lET Rn G0 H ;
1

where l is the latent heat of vaporization (J kg ), ET is evapotranspiration (kg m2 s1), Rn is net radiation (W m2; positive downward), G0 is soil surface heat ux (W m2; positive downward), and H is sensible heat ux (W m2; positive upward). The residual energy approach allows for estimation of lET by assuming that the net of energy uxes due to photosynthesis, respiration, and heat storage within the canopy are negligible ( o1% of incoming radiation; Meyers & Hollinger, 2004). Within each plot, observations were obtained via micrometeorological instrumentation wired to a datalogger (models CR10X or CR3000, Campbell Scientic Inc., Logan, UT, USA). The measurements of Rn, G0 and H, as described below,

Published 2010 This article is a US Government work and is in the public domain in the USA, GCB Bioenergy, doi: 10.1111/j.1757-1707.2010.01050.x

C O M P A R I S O N O F C A N O P Y E VA P O T R A N S P I R A T I O N
were made in 10-s intervals and averaged over 10-min intervals throughout the growing season. Measurements for analysis began before canopy closure, day of year (DOY) 137 for PRGs and DOY 168 for maize and collected continuously until senescence. Data loss from instrument failure occurred but represented o15% of total data points and, with only one exception, at least one plot was operational for all species at any given time period.

closure; therefore better characterization of G0 through larger numbers of subsamples would have minimal impacts on calculated lET (Fig. 2).

Sensible heat ux (H)

Sensible heat ux (H) was determined using a modied gradient ux method as: H pa cp Ts Ta ; ra 3

Net radiation (Rn)

Net radiation was measured using single-channel net radiometers (model Q*6 or Q*7; Radiation and Energy Balance Systems, Seattle, WA, USA) in each plot. Corrections were made for wind speed using a mechanical cup anemometer (model 12102D; R.M. Young Company, Traverse City, MI, USA) according to the manufacturers protocol. Net radiometers were positioned 4 m above the ground in the maize and miscanthus plots and 2.5 m above the ground in the switchgrass plots and remained at this height for the growing season. Before the growing season, all net radiometers were crosscalibrated against a factory calibrated model Q*7 net radiometer that was not deployed into the eld. Conditions where dew or precipitation interfered with data were excluded from the analysis. All net radiometers in the miscanthus plots were damaged during the last week of measurements. Values for Rn from switchgrass, which did not differ throughout the season from miscanthus, were substituted for this 1-week time period.

Soil heat ux (G0)

Soil heat ux (G0), is calculated as:   DT ; G0 G10 CDz Dt

where G10 is the soil heat ux at 10 mm soil depth, C is the volumetric heat capacity (MJ m3 1C1), and DT is the change in temperature in time increment Dt over soil depth Dz. One soil heat ux plate (model HFT-3; Radiation and Energy Balance Systems) was located in each plot, buried at a depth of 10 mm. Heat storage in the 10 mm of soil above each heat ux plate was obtained by placing a ne-wire thermocouple below the soil surface and above each heat ux plate. Since measurements of G0 were limited to one replicate per plot and given the small size of the heat ux plates (ca. 9 cm2) the measurement of this ux was less robust than Rn and H. However, G0 has been shown to have limited sensitivity on surface energy balance calculations (Kimball et al., 1994) particularly after canopy

where pa is air density (kg m3), cp the heat capacity of air (J kg1 1C1), Ts and Ta the surface and air temperatures ( 1C), and ra the aerodynamic resistance (s m1). Atmospheric resistance was modeled using different equations based on the wind speed, atmospheric stability, and canopy height (Jackson et al., 1987; Kimball et al., 1994, 1999; Triggs et al., 2004). Relative humidity was measured using a temperature and humidity probe (Model HMP-45C, Campbell Scientic Inc.) enclosed in an aspirated radiation shield (Model 076B Motor Aspirated Shield, MetOne Instruments Inc., Grants Pass, OR, USA) in a single location in a miscanthus plot at a height of 4.25 m. Wind speed was measured using a mechanical anemometer (Model 12102D 3-cup anemometer with photochopper, R.M. Young Company) and air temperature was measured using a thermistor (Model 107, Campbell Scientic Inc) with a custom aspirated heat shield using PVC pipe, an axial fan, and aluminum tape. These sensors were situated at a xed height aboveground to ensure that heights were no o1 m above the plant canopy at maximum height. This corresponded to a measured height, which varied among the plots, of between 2.25 and 2.35 m for switchgrass, 4.25 and 4.75 m for miscanthus, and 4.25 m for corn. A temperature probe malfunction in one of the miscanthus plots occurred around DOY 180, thus the air temperature reading for this plot was set equal to the closest miscanthus plots for the remainder of the growing season. Surface temperatures were measured using infrared radiometers (model IRR-P, Apogee Instruments, Logan, UT, USA). These sensors were mounted 4 m aboveground and positioned at 451 angles from horizontal facing south for all plots. When wind speed (u) was o0.1 m s1 during neutral conditions, dened as |TsTa|o0.1 1C, ra was set to a value of 1720 s m1 (Triggs et al., 2004). When u was o0.1 m s1 under nonneutral conditions, dened as |TsTa|40.1 1C, ra was calculated as (ASHRAE, 1972): ra pa cp 1:52jTs Ta j1=3 : 4

Published 2010 This article is a US Government work and is in the public domain in the USA, GCB Bioenergy, doi: 10.1111/j.1757-1707.2010.01050.x

4 G . C . H I C K M A N et al.
When u was higher than 0.1 m s1 the following equation was used:    1 1 z d zo 2 ra ln f; 5 u k zo rma ns constant (0.4), z is wind where k is Von Ka measurement height (m), zo and d are the roughness length (m; 0.1 canopy height) and zero plane displacement (0.65 canopy height), respectively (Campbell & Norman, 1998), and f is a stability correction. Height data, measured in weekly intervals for each plot, were tted to an equation that best described the measured heights throughout the growing season as in Bernacchi et al. (2007). While zo and d are likely to differ among species and growth stages, the values used in this paper are based on those suggested for uniform plant canopies (Campbell & Norman, 1998) and are assumed similar for all three species in this study as they are planted uniformly and are all considered tall grasses. A sensitivity analysis in which the values for zo and d were changed to 0.08 canopy height and 0.8 canopy height, respectively, resulted in less than a 1.5% change in either H or ET (data not shown). For stable conditions when the crop surface was cooler than air temperature, dened as TsoTa, f was calculated as (Mahrt & Ek, 1984): f 1 15Ri1 5Ri0:5 and for unstable conditions, Ts4Ta, as: ( ) 1 1 15Ri f ; 1 KRi0:5 6 et al., 2004). Previous studies have shown a radius of 12.5 m to be sufcient to support accurate measurements using this residual energy balance approach (Ottman et al., 2001). The residual energy balance approach to estimating lET is reasoned to have lower fetch requirements as turbulent transfer processes are a logarithmic function of height above the surface (Kimball et al., 1994). The residual energy balance method is most sensitive to measurements of Rn and Ts (Kimball et al., 1994). Since gradients are strongest near the crop surface, the layer near the canopy is most important for determining H (Kimball et al., 1999). The measurement of Ts using infrared radiometers minimized the fetch requirements relative to using a gradient ux approach where air temperature is measured at multiple heights. Similarly, given the sensor size and location within the plots, neither Rn nor Go are likely to be affected by the relatively small fetch. This leaves u and Ta as the two measurements that necessitate fetch considerations. Previous attempts to correct Ta to compensate for the smaller fetch of 12.5 m for free air CO2 enrichment (FACE) experiments resulted in negligible changes in the magnitude of Ta (Triggs et al., 2004). The fetch issues associated with measuring wind speed at 4 m above fragmented plant canopies are more difcult to correct for and likely add to the error associated with the measurements. The previous analyses of fetch issues focused primarily on FACE plots where the vegetation being measured is surrounded by the same species with the same canopy characteristics. In our study, the three species were surrounded on all sides by maize. Thus, the canopy height at maturity is similar, at least for maize and miscanthus, to the surrounding elds, thereby minimizing the gaps that might be introduced. However, a sensitivity analysis suggested the measurements most likely inuenced by the fetch constraints are small (Kimball et al., 1994) relative to the differences we observed for the species measured in this study.

where Ri is the Richardson number and is solved as Ri and K is solved as K 75k2 z d zo =zo 0:5 flnz d zo =zo g2 : 9 gTa Ts z d ; Ta 273:15u2 8

Biomass sampling
Under certain conditions, unreasonably high values for H were calculated at night. Thus, conditions when Rn was o10 W m2 and ra was calculated as o10 s m1, the data were excluded from analysis. As described in Dohleman & Long (2009), aboveground biomass was measured per unit land area on a biweekly basis. For miscanthus, 10 randomly selected tillers were taken from two randomly selected plants in each plot, and the total number of tillers on each plant was counted to scale the biomass of the sample to the individual plant. Samples were oven-dried at 75 1C until reaching a steady weight. Plant density was then used to scale sample weights to plot scale measurements. For switchgrass, the same protocol was followed, but the tiller samples from all three plots were mistakenly merged into one sample, thus removing

Fetch considerations
The 0.2 ha plots had a minimum functional fetch radius of only 15 m in the EW direction and 30 m in the NS direction which limited the possibility of using other micrometeorological techniques for directly measuring ET, as discussed previously (Kimball et al., 1999; Triggs

Published 2010 This article is a US Government work and is in the public domain in the USA, GCB Bioenergy, doi: 10.1111/j.1757-1707.2010.01050.x

C O M P A R I S O N O F C A N O P Y E VA P O T R A N S P I R A T I O N
the possibility to determine plot error. For maize, two subsamples per plot were removed and dried as before to determine dry sample mass. No samples of any species were taken within 3 m of the edge of a plot to avoid any border effects (Roberts et al., 1993). The last subsamples used in our analysis corresponded to the nal day of micrometeorological measurements after the canopies were completely senesced.

WUE
WUE was calculated from the total ET over the growing season and the biomass harvest that corresponded to the nal day of micrometeorological measurements which differed among the crops. The biomass components that were used for the WUE calculations are based on the biomass that was removed at harvest for the purpose of energy production. Biomass that would not be taken, including all belowground components, fallen residue, and portions of the tillers that were left standing after harvest, were not included. For corn, WUE is divided into two metrics; one similar to the PRGs where all aboveground biomass is considered and a second in which only the grain biomass is considered. The latter scenario is used as the current US ethanol production scenario where all ethanol production is based on grain fermentation.

Meteorology
For the purpose of describing the site meteorology, air temperature, relative humidity, and precipitation were collected from a weather station located in the center of the research eld. The Palmer Crop Moisture Index (PCMI), calculated from temperature, precipitation, and modeled soil water content, was used to estimate short-term moisture conditions (Palmer, 1968). PCMI values between 1 and 1 are indicative of normal moisture conditions, with values o1 suggesting more droughted conditions, and 41 suggesting overly moist conditions. The 30-year mean PCMI value for the region containing this site is ca. 0.4 (e.g., Bernacchi et al., 2007).

effect on the estimation of ET, as evaporation rarely occurs during precipitation, never occurs during dew, and because these conditions inuenced all three species in unison. Means of the lET were calculated for each species for the entire growing season to represent the total growing season water demand. All three species experienced a closed canopy from DOY 190 235, and this time period was analyzed separately to reect the inuences of physiological differences among the species. Mean values of latent heat ux (lET) for the respective time periods were summed and converted to ET (mm) using the latent heat of vaporization and the density of water. The difference in ET among species was statistically tested using a randomized complete block analysis of variance (PROC MIXED, GLM3 SAS 9.1, SAS Institute, Cary, NC, USA). Assumptions of normal distribution were veried using Levenes test of equality of variance and a proc univariate analysis for normality of data. Given the variability and low replicate number inherent within large eld experiments, a 5 0.10 was used in all cases to reduce the probability of Type II error. A Tukeys corrected a priori means separation was used to determine signicant differences between species. A comparison of regressions technique was used to test differences in accumulated biomass vs. accumulated ET (Mead & Curnow, 1983). This technique requires a polynomial that accurately reects the relationship between accumulated biomass vs. water use. A second-order polynomial was shown to t this criteria and was tted rst to all three species together and then to each species separately. The goodness-of-t for each separate curve was compared with the goodness-of-t for one line representing all data. The type 3 sum of squares (i.e., error variances around each model) and associated degrees of freedom for lumped vs. separate ts were used to compute F-ratios associated with tests of homogeneity of the tted polynomials (Mead & Curnow, 1983; Potvin et al., 1990). Signicant values for this test were set a priori at Po0.05.

Results

Data analysis
Micrometeorological data were collected in 10-s intervals and averaged over 10-min intervals throughout the growing season. Data points associated with instrument error or unfavorable meteorological conditions were removed from the dataset. These conditions were dened as periods when sensor accuracy was visibly diminished and were mostly associated with periods of heavy precipitation and occasionally with heavy dew formation. The removal of these data points has little

Meteorology/climatology
The 2007 growing season was slightly drier and warmer than the 30-year average as calculated by the Midwest Regional Climate Center (http://mrcc.sws.uiuc.edu/). Temperatures were ca. 1 1C warmer than the 30-year average for the year, and ca. 1.4 1C warmer during the growing season (DOY 140300). Annual and growing season precipitation were ca. 21% and ca. 22% less than the 30-year average, respectively. However, the PCMI for Illinois Climate Division 5, which includes the elds

Published 2010 This article is a US Government work and is in the public domain in the USA, GCB Bioenergy, doi: 10.1111/j.1757-1707.2010.01050.x

6 G . C . H I C K M A N et al.
in this study, only dropped below 1 once, for a period lasting ca. 7 days (Fig. 1). Solar radiation, wind, and other meteorological factors were within normal ranges of annual variation (Fig. 1).

Energy uxes
Two days from the growing season were selected to represent the diurnal trends for each of the measured energy uxes (Fig. 2). DOY 196 had very little cloud cover during the daylight hours, with no signicant differences in Rn observed among the three species (Fig. 2). Soil heat ux increased throughout the day until solar noon at which point it began to decline. This diurnal pattern was similar for all three species. Sensible heat ux was relatively steady throughout the day for miscanthus and maize, whereas switchgrass showed a more pronounced increase in H until midday. All three species had high rates of lET during the daylight hours, but the higher H for switchgrass resulted in a lower peak value of lET (Fig. 2). The overcast day, DOY 209, showed much more variability in Rn, H, and lET throughout the day and the differences among the species are much less (Fig. 2). Seasonal patterns in Rn were closely matched among the three species (Fig. 3). Day to day variability in daily mean Rn was observed as a consequence of meteorological conditions; however, no clear species effect in Rn was observed, either over the growing season as a whole or over the period in which the three canopies were closed (Fig. 3). Similarly, there are few differences in G0 among the three species (Fig. 3). There were observable differences in H among the three species, with switchgrass showing the highest H and miscanthus the lowest (Fig. 3). This was consistent across most of the growing season, except for after DOY 240 when H began to increase for maize. From this point on, maize had the highest H of the three species. The mean daily lET followed the inverse pattern of H (Fig. 3). When averaged over the growing season, miscanthus had the highest mean lET and there were no differences observed between maize and switchgrass (Table 1). The same pattern is apparent when considering the period of time in which all three canopies were closed (Table 1; Fig. 3). Despite the lack of statistically signicant differences between maize and switchgrass, there are lower rates of lET for switchgrass when the three species all experience mature, presenescent canopies (Fig. 3; Table 1).

Fig. 1 Daily maximum vapor pressure decit (VPD) and Palmer Crop Moisture Index (Palmer, 1968) (top panel), daily maximum and minimum air temperature (middle panel) and daily total solar radiation and precipitation (bottom panel) throughout the 2007 growing season.

ET and WUE
As determined from emergence to senescence for the PRGs and from planting to senescence for the maize, the

total growing season was longer for the PRGs by 33 days. Over their respective growing seasons, miscanthus had ca. 55% higher and switchgrass 25% higher ET compared with maize (Table 1). While the three species showed relatively similar water use during the period when the canopies were mature, the large disparity in water use between maize and the perennial species is attributed to the length of growing season (Fig. 4). Further, despite the difference in the growing season being only 33 days, it is clear that ET has started to decrease around 20 days before complete senescence for maize (Fig. 4). WUE based on aboveground-harvested biomass was lowest for switchgrass and highest for maize (Table 1). However, when only harvested grain biomass was used to calculate WUE, the value for maize declined by

Published 2010 This article is a US Government work and is in the public domain in the USA, GCB Bioenergy, doi: 10.1111/j.1757-1707.2010.01050.x

C O M P A R I S O N O F C A N O P Y E VA P O T R A N S P I R A T I O N

Fig. 2 Net radiation (Rn), soil heat ux (G0), sensible heat ux (H) and latent heat ux (lET) for a mostly sunny (left column, day 196) and an overcast (right column, day 209) day during the 2007 growing season. Open symbols represent switchgrass, grey symbols, maize and black symbols, miscanthus. Error bars represent 1 SE around the mean of three replicates.

almost one-third (Table 1). The amount of water used per unit of biomass accumulated differed among the three species based on a regression analysis (F2, 26 5 14.06, Po0.01), but these differences were driven by a strong deviation of switchgrass from the other two species (Fig. 5). The relationship between incremental biomass accumulation vs. ET was similar between maize and miscanthus (F1, 16 5 3.97, P40.05).

Discussion
This study tested the prediction that PRGs, which are likely to increase in percent coverage over the Midwestern US as the demand for cellulosic biomass increases, evapotranspire more than maize. It also tested whether the higher biomass production associated with miscanthus and switchgrass was sufcient to offset the

higher water use to result in an improved WUE compared with maize. Contrary to our predictions, switchgrass did not yield higher biomass than maize. Total season water use was 954 mm for miscanthus, 764 mm for switchgrass, and 611 mm for maize, supporting the hypothesis that the PRGs will use more water over the duration of the growing season. Despite the higher productivity of miscanthus, WUE at ca. 19 kg ha1 mm1 was similar to maize at ca. 18.6 kg ha1 mm1 (Table 1). Switchgrass used more water than maize but had lower harvested biomass; thus WUE for switchgrass was about half of the values measured for the other two species. The 2007 growing season was drier than the 30-year average; however, at only one point in the growing season did the Palmer Crop Moisture Index drop below 1 (a low of 1.1 was achieved around

Published 2010 This article is a US Government work and is in the public domain in the USA, GCB Bioenergy, doi: 10.1111/j.1757-1707.2010.01050.x

8 G . C . H I C K M A N et al.

Fig. 3 Daily mean values of net radiation (Rn), soil heat ux (G0), sensible heat ux (H) and latent heat ux (lET) over the 2007 growing season for switchgrass, maize and miscanthus. The left panels show the entire growing season and the right panels show an expanded view of the time increment during which the canopies for all three species were closed. Symbols and error bars are as in Fig. 2.

Table 1 Latent heat ux (lET), cumulative water use and water-use efciency (WUE) for Miscanthus giganteus (miscanthus), Panicum virgatum (switchgrass) and Zea mays (maize) estimated using a residual energy balance approach Seasonal values Miscanthus Switchgrass Maize Maize grain Mature canopy (DOY 190235) Miscanthus Switchgrass Maize Elapsed days 166 166 133 133 45 45 45 Mean lET (W m2) 159.7 127.9 127.8 127.8 2.8a 5.6b 4.8b 4.8b Water use (mm) 954.6 764.3 611.9 611.9 16.4a 33.7b 23.2c 23.2c WUE (kg ha1 mm1) 19.1 9.7 29.7 18.6 0.3a 0.4b 1.1c 0.8a

171.4 0.2a 144.8 6.6b 150.6 3.7b

277.7 0.4a 234.6 10.7b 243.9 5.9b

Biomass data used in the calculations of WUE for miscanthus and maize is from Dohleman & Long (2009) and for switchgrass is from F. G. Dohleman et al. (unpublished results). Values with differing superscript letters signify statistical differences at Po0.05. ET, evapotranspiration; DOY, day of the year.

Published 2010 This article is a US Government work and is in the public domain in the USA, GCB Bioenergy, doi: 10.1111/j.1757-1707.2010.01050.x

C O M P A R I S O N O F C A N O P Y E VA P O T R A N S P I R A T I O N

Fig. 4 Contour maps showing latent heat ux (W m2) over the diurnal time course (y-axis) throughout the 2007 growing season (x-axis) for miscanthus (top), switchgrass (middle) and maize (bottom).

Fig. 5 Aboveground biomass as a function of water use during the 2007 growing season for miscanthus, switchgrass and maize. For each point, the total water used was determined according to the time in which a biomass harvest was performed. Data are only presented for presenesced biomass harvests. Biomass data are from Dohleman & Long (2009) for miscanthus and maize and from F. G. Dohleman (unpublished results) for switchgrass. The line represents a second-order polynomial tted to the miscanthus and maize data and symbols and error bars are as in Fig. 2.

mid-June) signifying abnormally dry conditions (Fig. 1). Throughout the rest of the growing season PCMI values fell within the Slightly Dry/Favorably Moist category. Thus, although this season was slightly drier than normal, there is no indication of drought conditions or plant stress e.g., low yields or low harvest percentages (Dohleman & Long, 2009). The perennial nature of miscanthus and switchgrass results in a longer growing season and more opportunity to evapotranspire relative to maize. Mean lET was 25% higher for miscanthus relative to maize and dropped to only about 18% higher than maize when considering the time period when both canopies were closed (Table 1). Thus, for miscanthus, there is not only the effect of a longer growing season, but also physiological or morphological factors that drive increased ET. In contrast, switchgrass mean seasonal and mean mature canopy lET was not statistically different from maize (Table 1). Given the small difference in mean lET, the higher ET for switchgrass relative to maize is likely dominated by the longer growing season (Fig. 4). These results suggest that a combination of physiological and morphological factors, in addition to growing season length, drive the observed differences among the species. Potential morphological differences include mean root mass and distribution (Neukirchen et al., 1999), LAI

and aboveground biomass (Heaton et al., 2008; Dohleman & Long, 2009). Differences among the species in stomatal conductance, which is shown to be closely coupled to canopy water use (e.g., Bernacchi et al., 2007), could also drive the differences in lET. During the times when the PRGs are evapotranspiring and the maize eld is fallow some evaporation in maize will occur. While this component of ET was not considered here, rates of evaporation outside of the growing season are shown to be quite small relative to what occurs during the growing season. For example, a recent study shows that ca. 80% of the annual ET in a maize/soybean ecosystem occurs during the three to four month growing season (Suyker & Verma, 2009). The remaining 20% is evapotranspired during the remaining months, which, when considering the difference in growing season lengths between the PRGs and maize, amounts to a very small percentage. Measurements of lET over the fallow maize eld were not measured in this study. However, when the growing season for maize was extrapolated to the same length as the perennial grasses, the total amount of ET for maize increased to 763 mm, similar to the amount of water evapotranspired from switchgrass. It is unlikely, however, that the evaporation from a fallow maize eld could account for the large differences observed between maize and miscanthus. When assessed over the entire growing season and using total aboveground biomass, WUE was lowest for

Published 2010 This article is a US Government work and is in the public domain in the USA, GCB Bioenergy, doi: 10.1111/j.1757-1707.2010.01050.x

10 G . C . H I C K M A N et al.
switchgrass and highest for maize. The difference in WUE between maize and miscanthus is minimized when, instead of using the total aboveground biomass, only the grain component of maize is used in calculating WUE (Table 1). Since residue removal is known to have negative implications for sustainability of maize (Mann et al., 2002; Andrews, 2006; Blanco-Canqui & Lal, 2009), there exists a trade-off between maximizing biomass harvest to increase WUE and agricultural sustainability. Since the purpose of these ecosystems are for the sustainable production of feedstock for renewable energy production, the corn-grain WUE scenarios is a more relevant comparison with the perennial species. The relationship between biomass accumulation and ET was similar between miscanthus and maize (Fig. 5), but why does the end of season WUE between these species differ (Table 1)? This disparity can be partially explained through losses in aboveground biomass for both species after senescence, but to a greater extent for miscanthus (Dohleman & Long, 2009). The loss in standing biomass immediately following senescence for miscanthus appears to be exceptional in this measurement season (Clifton-Brown et al., 2001; Heaton et al., 2008; Dohleman & Long, 2009). Similarly, the loss in aboveground biomass for maize suggested a harvest index of 0.66, which is much higher than typically measured (e.g., Tollenaar et al., 2006). Therefore, these results may underestimate the total seasonal WUE for both miscanthus and maize. The values for WUE calculated from total season ET and corn grain biomass correspond well with the range of values reported by a number of previous studies conducted within several management practices and climate regimes (Hateld et al., 2001; Zwart & Bastiaanssen, 2004; Suyker & Verma, 2009). Mean WUE averaged over the globe was estimated at 18 6.9 kg ha1 mm1 (Zwart & Bastiaanssen, 2004) which is remarkably similar to the value obtained in this study (18.6 kg ha1 mm1). Similarly, a study conducted in Nebraska, USA calculated WUE as 20 kg ha1 mm1 (Suyker & Verma, 2009). The differences among these three species water use revealed valuable insights to the potential environmental impacts of second-generation bioenergy crop growth. Given the demonstrated increases in ET associated with the growth of second-generation bioenergy crops both temporally and quantitatively, we can expect changes in the environmental services tied to the hydrological cycle (Stohlgren et al., 1998; Brown et al., 2005; Li et al., 2007). Our study demonstrated increased seasonal ET associated with the growth of second-generation bioenergy crops driven by a combination of physiological, morphological, and phenological differences among these three species. Therefore, potential largescale plantings of bioenergy crops could increase ET, thereby decreasing surface temperatures (Georgescu et al., 2009), increasing humidity, increasing precipitation and cloud cover, and decreasing solar radiation (Sellers et al., 1997; Bounoua et al., 2002; Findell et al., 2007). Increased ET associated with PRGs could also impact runoff and groundwater quality/quantity, soil moisture, and river ow (Stohlgren et al., 1998; Brown et al., 2005; Li et al., 2007). The primary impacts of altered ET (e.g. soil moisture) would also affect secondary changes, such as rates of carbon sequestration or soil respiration (Pastor & Post, 1986; Ma et al., 2000), and soil nitrogen budgets (Pastor & Post, 1986). Depending on the fraction coverage of PRGs grown in a region, the important link between vegetation and hydrology should be a major consideration in sustainable biomass production.

Acknowledgements
The authors wish to thank Don Ort, David Drag, Marcelo Zeri, Ursula Melissa Ruiz Vera, Christina Burke, Sarah Campbell, and Greg McIsaac for helpful comments on the manuscript. We also wish to acknowledge Devin Quarles for assisting with the eld measurements. This research was supported by the Energy Biosciences Institute.

References
American Society of Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE) (1972) ASHRAE Handbook of Fundamentals. ASHRAE, New York, NY, USA. Andrews SS (2006) Crop Residue Removal for Biomass Energy Production: Effects on Soils and Recommendations pp. 115. USDA-Natural Resource Conservation Service, Washington, DC, USA. Bernacchi CJ, Kimball BA, Quarles DR, Long SP, Ort DR (2007) Decreases in stomatal conductance of soybean under open-air elevation of CO2 are closely coupled with decreases in ecosystem evapotranspiration. Plant Physiology, 143, 134144. Blanco-Canqui H, Lal R (2009) Crop residue removal impacts on soil productivity and environmental quality. Critical Reviews in Plant Sciences, 28, 139163. Bounoua L, Defries R, Collatz GJ, Sellers P, Khan H (2002) Effects of land cover conversion on surface climate. Climatic Change, 52, 2964. Brown AE, Zhang L, McMahon TA, Western AW, Vertessy RA (2005) A review of paired catchment studies for determining changes in water yield resulting from alterations in vegetation. Journal of Hydrology, 310, 2861. Campbell GS, Norman JM (1998) An Introduction to Environmental Biophysics, 2nd edn. Springer-Verlag, New York. Clifton-Brown JC, Lewandowski I, Andersson V et al. (2001) Performance of 15 Miscanthus genotypes at ve sites in Europe. Agronomy Journal, 93, 10131019.

Published 2010 This article is a US Government work and is in the public domain in the USA, GCB Bioenergy, doi: 10.1111/j.1757-1707.2010.01050.x

C O M P A R I S O N O F C A N O P Y E VA P O T R A N S P I R A T I O N
De Fraiture C, Giordano M, Liao Y (2008) Biofuels and implications for agricultural water use: blue impacts of green energy. Water Policy, 10, 6781. Dohleman FG, Long SP (2009) More productive than maize in the Midwest: how does miscanthus do it? Plant Physiology, 150, 21042115. Farrell AE, Plevin RJ, Turner BT, Jones AD, OHare M, Kammen DM (2006) Ethanol can contribute to energy and environmental goals. Science, 311, 506508. Findell KL, Shevliakova E, Milly PCD, Stouffer RJ (2007) Modeled impact of anthropogenic land cover change on climate. American Meteorological Society, 20, 36213634. Georgescu M, Lobell DB, Field CB (2009) Potential impact of U.S. biofuels on regional climate. Geophysical Research Letters, 36 L21806. Gurgel A, Reilly JM, Paltsev S (2007) Potential land use implications of a global biofuels industry. Journal of Agricultural and Food Industrial Organization, 5, 134. Hall RL (2003) Grasses for Energy Production. Hydrological Guidelines, pp. 115. Department of Trade and Industry, New and Renewable Energy Programme Centre for Ecology and Hydrology, London, UK. Hamelinck CN, Hooijdonk Gv, Faaij AP (2005) Ethanol from lignocellulosic biomass: techno-economic performance in short-, middle- and long-term. Biomass and Bioenergy, 28, 384410. Hateld JL, Sauer TJ, Prueger J (2001) Managing soils to achieve greater water use efciency: a review. Agronomy Journal, 93, 271280. Heaton EA, Dohleman FG, Long SP (2008) Meeting US biofuel goals with less land: the potential of Miscanthus. Global Change Biology, 14, 20002014. Heaton EA, Long SP, Voigt TB, Jones MB, Clifton-Brown J (2004) Miscanthus for renewable energy generation: European Union experience and projections for Illinois. Mitigation and Adaptation Strategies for Global Change, 9, 433451. Huband NDS, Monteith JL (1986) Radiative surface temperature and energy balance of a wheat canopy. Boundary-Layer Meteorology, 36, 117. Jackson RD, Moran MS, Gay LW, Raymond LH (1987) Evaluating evaporation from eld crops using airborne radiometry and ground-based meteorological data. Irrigation Science, 8, 8190. Kimball BA, LaMorte RL, Pinter PJ Jr et al. (1999) Free-air CO2 enrichment and soil nitrogen effects on energy balance and evapotranspiration of wheat. Water Resources Research, 124, 11791190. Kimball BA, LaMorte RL, Seay RS et al. (1994) Effects of free-air CO2 enrichment on energy balance and evapotranspiration of cotton. Agricultural and Forest Meteorology, 70, 259278. Lewandowski I, Scurlock JMO, Lindvall E, Christou M (2003) The development and current status of perennial rhizomatous grasses as energy crops in the US and Europe. Biomass and Bioenergy, 25, 335361. Li KY, Coe MT, Ramankutty N, Jong RD (2007) Modeling the hydrological impact of land-use change in West Africa. Journal of Hydrology, 337, 258268. Ma Z, Wood CW, Bransby DI (2000) Soil management impacts on soil carbon sequestration by switchgrass. Biomass and Bioenergy, 18, 469477.

11

Mahrt L, Ek M (1984) The inuence of atmospheric stability on potential evaporation. Journal of Climate and Applied Meteorology, 23, 222234. Mann L, Tolbert V, Cushman J (2002) Potential environmental effects of corn (Zea mays L.) stover removal with emphasis on soil organic matter and erosion. Agriculture, Ecosystems and Environment, 89, 149166. McLaughlin SB, Walsh ME (1999) Evaluating environmental consequences of producing herbaceous crops for bioenergy. Biomass and Bioenergy, 14, 317324. Mead R, Curnow RN (1983) Statistical Methods in Agriculture and Experimental Biology. Chapman & Hall, London. Meyers TP, Hollinger SE (2004) An assessment of storage terms in the surface energy balance of maize and soybean. Agricultural and Forest Meteorology, 125, 105115. Midwest Regional Climate Center (MRCC) (2009) Climate of the Midwest. Available at: http://mrcc.isws.illinois.edu (accessed 12 October 2008). Msangi S, Sulser T, Rosegrant M, ValmonteSantos R, Ringleri C (2006) Global Scenarios for Biofuels: Impacts and Implications, pp. 116. International Food Policy Research Institute (IFPRI) Washington, DC, USA. Neukirchen D, Himken M, Lammel J, Czypionka-Krause U, Olfs H-W (1999) Spatial and temporal distribution of the root system and root nutrient content of an established Miscanthus crop. European Journal of Agronomy, 11, 301309. Ottman MJ, Kimball BA, Pinter PJ et al. (2001) Elevated CO2 increases sorghum biomass under drought conditions. New Phytologist, 150, 261273. Palmer WC (1968) Keeping track of crop moisture conditions, nationwide: the new Crop Moisture Index. Weatherwise, 21, 156161. Pastor J, Post WM (1986) Inuence of climate, soil moisture, and succession on forest carbon and nitrogen cycles. Biogeochemistry, 2, 327. Potvin C, Lechowicz MJ, Tardif S (1990) The statistical analysis of ecophysiological response curves obtained from experiments involving repeated measures. Ecology, 71, 13891400. Roberts MJ, Long SP, Tieszen LL, Beadle CL (1993) Measurement of plant biomass and net primary production of herbaceous vegetation. In: Photosynthesis and Production in a Changing Environment: A Field and Laboratory Manual (eds Hall DO, Scurlock JMO, Bolhar-Nordenkampf HR, Leegood RC, Long SP), pp. 121. Chapman & Hall, London. Sanderson MA, Adler PR (2008) Perennial forages as second generation bioenergy crops. International Journal of Molecular Sciences, 9, 768788. Searchinger T, Heimlich R, Houghton RA et al. (2008) Use of US croplands for biofuels increases greenhouse gases through emissions from land-use change. Science, 319, 12381240. Sellers PJ, Dickinson RE, Randall DA et al. (1997) Modeling the exchanges of energy, water, and carbon between continents and the atmosphere. Science, 275, 502509. Sims RE, Hastings A, Schlamadinger B (2006) Energy crops: current status and future prospects. Global Change Biology, 12, 20542056. Stephens W, Hess T, Knox JW (2001) Review of the Effects of Energy Crops on Hydrology. Institute of Water and Environment, Craneld University, Silsoe.

Published 2010 This article is a US Government work and is in the public domain in the USA, GCB Bioenergy, doi: 10.1111/j.1757-1707.2010.01050.x

12 G . C . H I C K M A N et al.
Stohlgren TJ, Chase TN, Pielke RAS, Kittel TGF, Baron JS (1998) Evidence that local land use practices inuence regional climate, vegetation, and stream ow patterns in adjacent natural areas. Global Change Biology, 4, 495504. Suyker AE, Verma SB (2009) Evapotranspiration of irrigated and rain-fed maize-soybean cropping systems. Agricultural and Forest Meteorology, 149, 443452. Tan KT, Lee KT, Mohamed AR (2008) Role of energy policy in renewable energy accomplishment: the case of second-generation bioethanol. Energy Policy, 36, 33603365. Tollenaar M, Deen W, Echarte L, Liu W (2006) Effect of crowding stress on dry matter accumulation and harvest index in maize. Agronomy Journal, 98, 930937. Triggs JM, Kimball BA, Pinter PJ Jr et al. (2004) Free-air CO2 enrichment effects on the energy balance and evapotranspiration of sorghum. Agricultural and Forest Meteorology, 124, 6379. Zwart SJ, Bastiaanssen WGM (2004) Review of measured crop water productivity values for irrigated wheat, rice, cotton and maize. Agricultural Water Management, 69, 115133.

Published 2010 This article is a US Government work and is in the public domain in the USA, GCB Bioenergy, doi: 10.1111/j.1757-1707.2010.01050.x